This application proposes advances to precision medicine along three complementary arcs, building on our initial work in RO1-MH099003 (Characterizing Placebo Response to Active Treatment Using Very High-Dimension- al Data). A primary component of precision medicine involves determining optimal treatment decisions using baseline variables. We have previously developed statistical models that can accommodate both scalar and functional predictors in determining treatment decision functions. This renewal application addresses the fact that available data modalities continue to evolve and grow ever more complex. Determining treatment deci- sions is closely related to the problem of quantifying the degree to which individual outcomes are due to spe- cific drug effects versus placebo responses. We propose to develop completely nonparametric distance-based approaches with the goal of determining optimal treatment decisions and generalizing these to accommodate multiple data modalities. Because no single biomarker for response to treatment has been identified, methods to form composite biomarkers by combining predictors will be developed to incorporate data arising from com- plex modalities. Many brain-related measures are expensive and time consuming to collect, and thus a vital component of optimizing treatment decisions is to develop a sequential treatment decision strategy whereby easy-to-make measurements are obtained first, and the more expensive and time consuming measurements are only obtained if they are needed to improve confidence in a treatment decision. A complementary arc of the proposed research involves stratified psychiatry and developing directed partitioning/clustering methods to improve precision medicine. This is not only important to the work on deter- mining optimal treatment decisions, but also towards a data-driven discovery of psychiatric biotypes. Optimizing treatment decisions is naturally related to the thorny problem of making a clear diagnosis, especially in mental health where there is a wide degree of heterogeneity and overlap in current disease phenotypes.
This research aims to expand the notion of endophenotypes from genetics to data modalities based on brain structure, func- tion, and integrity. Specific to this application, the notion of endophenotypes will be expanded to discover more stable phenotypes using biomedical technologies, that those obtained based on symptoms. The Establishing Moderators & Biosignatures for Antidepressant Response in Clinical Care (EMBARC; U01MH092221, U01MH092250 ) study, which is the most ambitious systematic effort to discover biomarkers to guide treatment of major depressive disorder, has collected and made publicly available an unprecedented collection of clinical and biological patient phenotypes. This is a proposal to develop analytic methodology that would be able to take full advantage of rich and complex patient data as those in the EMBARC study for the purposes of precision medicine, in an anticipation of future similar or even more ambitious studies.
Utilizing biologically-based measures from high-dimensional and multi-modal data sources such as brain im- ages, in addition to traditional clinical measures, appears to be the most likely path to substantial progress towards determining optimal patient-specific treatment decisions for psychiatric illnesses. Functional statistical methods, coupled with innovative clustering approaches, will be developed to fully exploit the power of multiple modern data platforms. The statistical framework for treatment classification from biological brain-based mea- sures shares common characteristics to classification of psychiatric illnesses and developed methods can be applied to psychiatric nosology.
|Ciarleglio, Adam; Petkova, Eva; Ogden, Todd et al. (2018) Constructing treatment decision rules based on scalar and functional predictors when moderators of treatment effect are unknown. J R Stat Soc Ser C Appl Stat 67:1331-1356|
|Petkova, Eva; Ogden, R Todd; Tarpey, Thaddeus et al. (2017) Statistical Analysis Plan for Stage 1 EMBARC (Establishing Moderators and Biosignatures of Antidepressant Response for Clinical Care) Study. Contemp Clin Trials Commun 6:22-30|
|Jiang, Bei; Petkova, Eva; Tarpey, Thaddeus et al. (2017) LATENT CLASS MODELING USING MATRIX COVARIATES WITH APPLICATION TO IDENTIFYING EARLY PLACEBO RESPONDERS BASED ON EEG SIGNALS. Ann Appl Stat 11:1513-1536|
|Reiss, Philip T; Goldsmith, Jeff; Shang, Han Lin et al. (2017) Methods for scalar-on-function regression. Int Stat Rev 85:228-249|
|Petkova, Eva; Tarpey, Thaddeus; Su, Zhe et al. (2017) Generated effect modifiers (GEM's) in randomized clinical trials. Biostatistics 18:105-118|
|Cloitre, Marylene; Petkova, Eva; Su, Zhe et al. (2016) Patient characteristics as a moderator of post-traumatic stress disorder treatment outcome: combining symptom burden and strengths. BJPsych Open 2:101-106|
|Ciarleglio, Adam; Petkova, Eva; Tarpey, Thaddeus et al. (2016) Flexible functional regression methods for estimating individualized treatment regimes. Stat (Int Stat Inst) 5:185-199|
|Tarpey, Thaddeus; Petkova, Eva; Zhu, Liangyu (2016) Stratified Psychiatry via Convexity-Based Clustering with Applications Towards Moderator Analysis. Stat Interface 9:255-266|
|Ciarleglio, Adam; Petkova, Eva; Ogden, R Todd et al. (2015) Treatment decisions based on scalar and functional baseline covariates. Biometrics 71:884-94|
|Tarpey, Thaddeus; Ogden, R Todd; Petkova, Eva et al. (2015) Reply Am Stat 69:254-255|
Showing the most recent 10 out of 11 publications